胎盘超声图像自动分级的初步探索

摘要

目的 建立自动提取胎盘超声图像特征并对其自动分级的方法.方法 首先对142例正常妊娠妇女的胎盘进行常规超声分级,再随机抽取其中69例胎盘图像,提取图像的纹理特征,利用模糊分类法建立自动分级的计算系统;另73例用于测试、比较两种方法对胎盘分级的差别.结果 模糊分类法与常规Grannum胎盘分级相比,在判定胎盘分级上平均误差为0.21级(P>0.05),对Ⅰ、Ⅱ、Ⅲ级胎盘误差分别为0.13、0.35和0.20级(P均>0.05).结论 应用胎盘超声图像自动分析方法可对胎盘分级进行客观、可重复性的评价,在临床上有一定推广价值.%Objective To establish an automatic grading method by abstracting placenta texture features.Methods The regular Grarmum placenta grading method applied was first in 142 normal pregnant women,then fuzzy classification method were used to dispose 69 randomly acquired placenta pictures for getting characteristic parameters to establish the new method.Seventy-three placenta pictures were adapted to compare the regular method with the automatic one.Results These two methods had average error of 0.21(P>0.05)and for grade Ⅰ、Ⅱ、Ⅲ placenta,the average errors were 0.13,0.35,0.20,respectively(all P>0.05).Conclusions The fuzzy classification method applying in placenta grading is an objective and highly reproductive method.